Background:

• Rationale: Although local and cytokine-mediated inflammation may help clear bacterial pathogens, this
inflammation also exacerbates pulmonary dysfunction by impairing alveolar gas exchange. Furthermore,
severe systemic inflammation can contribute to sepsis and end-organ dysfunction. Systemic
corticosteroids may attenuate the inflammatory response and thus reduce ARDS, clinical instability, etc.
o MoA: blockage of promoter sites of pro-inflammatory genes (e.g. IL-1ß), recruitment of
transcription factors to anti-inflammatory genes (e.g. IL-10), inhibition of almost all
inflammatory cytokines via blockage of NF-kB and AP-1
• My Interpretation: plausible biological hypothesis

Methods:
• Eligible studies satisfy three criteria:
o i) RCT in which adults w/ CAP are assigned to IV/PO steroid or placebo/nothing
o ii) Report on ≥1 of: duration of hospitalization, time to clinical stability, all-cause mortality,
need for mechanical ventilation, need for ICU admission, ARDS development
o iii) No exclusion criteria: VAP, aspiration pneumonia, PCP pneumonia, COPD only
§ **Effective additional exclusion criteria due to underlying studies
• Search Methodology: replicates prior Cochrane review on the topic—2 pairs of 2 reviewers search
“pneumonia” and “corticosteroid” in MEDLINE, EMBASE, and Cochrane registry; included references
o Given prior review, authors limited search to studies between 1/1/2010 and 5/24/2015
• Data Abstraction & Quality Assessment:
o Reviewers extracted data on possible harms: re-hospitalization, hyperglycemia requiring
treatment, GI hemorrhage, severe neuropsychiatric symptoms (delirium, psychosis, mania)
o Reviewers extracted data on pneumonia severity: PSI (IV/V) > CURB65 (≥2) > IDSA/ATS (1
major or 3minor) > other
§ Studies were designated “severe illness” if ≥70% had severe illness or if ≥15% of
control population died
o GRADE System: well-accepted method of rating quality of evidence underpinning clinical
recommendations from systematic reviews—quality may be very low, low, moderate, or high
§ i) Risk of Bias: assessed via modified Cochrane instrument
• Inherent limitations in design/conduct of a study
§ ii) Imprecision: optimal information size calculations (alpha = 0.05, beta = 0.80)
• Certainty is lower when clinical decision would be different depending on
location of true effect within the confidence interval
§ iii) Inconsistency: unclear how this was assessed
• Several studies show consistent effects
§ iv) Indirectness: assessed subjectively based on disease criteria used in each study
• Studies directly compare interventions of interest in populations of interest and
report outcomes critical for decision making
§ v) Publication Bias: assessed via visual inspection of funnel plots
• Extent to which outcome of study influences decision to publish it
§ Other: accounted for effect of loss due to follow up
• Data Synthesis & Analysis:
o Random-effects model for all analysis: assumes data being analyzed are drawn from a hierarchy
of different populations whose differences relate to their place in that hierarchy
§ Hypothesized the following would be associated with larger treatment benefit:
publication before 2000, greater CAP severity, longer steroid tx (>3 v. ≤3d), higher bias
• My Interpretation: well-defined albeit narrow eligibility, excellent search methodology, excellent
characterization of underlying study quality, reasonable data analysis method
Results:
• Literature Search: 3281 unique citations + 3 studies from references à 13 RCTs (2005 patients)
o Mostly European, all but one without pharma funding
o Sample sizes 30-784, ~60% men, median age early 60s
o Steroids: dexamethasone, prednisone, prednisolone, methylprednisolone, hydrocortisone
§ Doses were variable, duration of treatment ranged from 1 dose to 10d
o Follow up: ranged from in-hospital to 60d from enrollment
o Largely excluded the following: GI hemorrhage within 3mo (7 studies), immunosuppression (11
studies), pregnant women (6 studies)
• Risk of Bias: 5/13 trials (~70.4% of patients) had low risk of bias. Insignificant loss to follow up.
o Could only assess all-cause mortality for publication bias (!)
• Outcomes:
o I) All-Cause Mortality: 12 trials
§ Population Results: RR 0.67 [95% CI, 0.45-1.01
• Corticosteroid: 52/977 (5.3%) died
• Control: 79/997 (7.9%) died
§ Subgroup Results:
• Severe Pneumonia (6 trials, 388 patients): RR 0.39 [CI, 0.20-0.77]
• Less Severe (6 trials, 1586 patients): RR 1.00 [CI, 0.79-1.26]
§ Moderate certainty given CI crossing 1 and possible subgroup effect
o II) Mechanical Ventilation: 5 trials
§ Population Results: RR 0.45 [CI, 0.26-0.79]
• Corticosteroid: 17/550 (3.1%) got ventilated
• Control: 29/510 (5.7%) got ventilated
§ Subgroup results:
• Severe Pneumonia (3 trials, 230 patients): RR 0.54 [CI, 0.50-0.58]
• Less Severe (2 trials, 830 patients): RR 0.18 [CI, 0.08-0.43]
§ Moderate certainty given small number of events (46 total)
o III) ICU Admission: 3 trials, all “less severe,” all steroids >3d
§ Population results: RR 0.69 [CI, 0.46-1.03]
o IV) ARDS: 4 trials (one didn’t specify diagnostic criteria)
§ Population Results: RR 0.24 [CI, 0.10-0.56]
o V) Duration of Hospitalization: 9 studies (all discharged at discretion of admitting physician)
§ Population Results: -2.96d [CI, -5.18d – -0.75d]
§ Subgroup Results:
• Low risk of bias (3 trials, 1288 patients): -1.00d [CI, -1.79d – -0.21d]
• High risk of bias (6 trials, 356 patients): -4.41d [CI, -7.65d – -0.75d]
o VI) Time to Clinical Stability: 5 trials (stable VS, no O2 requirement)
§ Population Results: -1.22d [CI, -2.08 – -0.35d]
o VII) Adverse Effects:
§ Hyperglycemia requiring treatment (6 studies, 1534 patients): RR 1.49 [CI, 1.01-2.19]
§ GI Hemorrhage (7 studies, 1223 patients), severe neuropsychiatric (4 studies, 1217
patients), and re-hospitalization (2 studies, 1089 patients) showed no significant efffect
• My Interpretation:
o Literature Search: increasingly narrow eligibility, unclear steroid dosing
o Risk of Bias: underestimated risk of publication bias
o Outcomes:
§ Inverted relationship between all-cause mortality subgroup analysis and mechanical
ventilation subgroup analysis suggests that neither effect is probably entirely real
§ Convincing evidence for some benefit in mechanical ventilation, ARDS, duration of
hospitalization, time to clinical stability
 § Definite correlation between steroids and hyperglycemia, but not other more severe AEs!

Discussion:
• Summary of important findings:
o Moderate certainty: 5% reduction in need for mechanical ventilation & rate of ARDS (NNT 20)
o High Certainty: reduced time to clinical stability & duration of hospitalization by 1d
o High Certainty: increased incidence of hyperglycemia requiring treatment
o Equivocal: reduction in mortality depends on subgroup analysis
§ Pros: large magnitude of effect, biological plausibility, small interaction P-value
§ Cons: based on differences between rather than within studies, driven by small study that
was stopped early, not consistent with other subgroup effects (ventilation, ARDS)
• Severe Pneumonia Mortality with and without reference 24:
o With: 16/215 (7.4%) steroid mortality v. 38/173 (22.0%) control mortality
o Without: 16/192 (8.3)% steroid mortality v. 30/152 (19.7%) control
mortality
• Less Severe Pneumonia Mortality:
o 36/762 (4.7%) steroid mortality v. 41/824 (5.0%) control mortality
• Author’s Assessment of Strengths of Study: explicit eligibility criteria, comprehensive search, bias
assessment, GRADE criteria
• Author’s Assessment of Limitations of Study: unclear steroid dosing/ROA, small number of events
(ventilation, ICU admissions, ARDS)

Conclusion
• Given plausible biological mechanism, high likelihood of certain benefits (mechanical ventilation,
ARDS, hospital duration), relatively unimportant adverse effects (hyperglycemia), and the cost and
incidence associated with CAP, it seems reasonable to apply glucocorticoids in adult patients in their 60s
admitted to hospital with CAP who do not fall under any of the exclusion criteria:
o Exclusion: VAP, aspiration pneumonia, PCP pneumonia, COPD only, pregnant, recent GI bleed
within 3mo, immunosuppression
• Appreciate that this trial probably has not demonstrated unequivocal benefit of corticosteroids, but that it
certainly suggests it and that its evidence is credible. Consider the following rebuttal from Wan-Jie Gu
• Consider UpToDate’s recommendation regarding corticosteroids in CAP:
o For patients with CAP who have evidence of an exaggerated or dysregulated host inflammatory
response, defined as sepsis or respiratory failure with an FiO2 requirement of >50 percent
plus one or more of the following features (metabolic acidosis with an arterial pH of <7.3,
lactate >4 mmol/L, or a C-reactive protein >150 mg/L), we suggest giving adjunctive
glucocorticoids. These patients are at high risk of mortality and are likely to benefit the most.
o Reasons to avoid glucocorticoids in such patients include risk factors for severe adverse events
such as recent gastrointestinal bleeding, poorly controlled diabetes, or severe
immunocompromise. We also avoid glucocorticoids in patients with CAP known to be caused by
a viral pathogen such as influenza or a fungal pathogen such as Aspergillus.
o When using adjunctive glucocorticoids, we treat for five days:
§ For patients who are unable to take oral medications, we
use methylprednisolone 0.5 mg/kg IV every 12 hours.
§ For patients who can take oral medications, we use prednisone 50 mg orally daily.
• Important questions in the future:
o Are glucocorticoids beneficial in the highest risk patients? Especially immunocompromised,
pregnant, patients at risk of delirium, etc.